SwinIR VS meadowrun

We’re referencing our three models directly as git repos (because they’re not available as packages in PyPI) in model_requirements.txt, but we had to make a few changes to make these repos work as pip packages. Pip looks for a setup.py file in the git repo to figure out which files from the repo need to be installed into the environment, as well as what the dependencies of that repo are. GLID-3-xl and latent-diffusion (another diffusion model that GLID-3-xl depends on) had setup.py files that needed tweaks to include all of the code needed to run the models. SwinIR didn’t have a setup.py file at all, so we added one. Finally, all of these setup.py files needed additional dependencies, which we just added to the model_requirements.txt file.

If you’re anything like us, though, you’ll feel compelled to poke around the code and run the model yourself. We’ll do that in this article using Meadowrun, an open-source library that makes it easy to run Python code in the cloud. For ML models in particular, we just added a feature for requesting GPU machines in a recent release. We’ll also feed the images generated by DALL·E Mini into additional image processing models (GLID-3-xl and SwinIR) to improve the quality of our generated images. Along the way we’ll deal with the speedbumps that come up when running open-source ML models on EC2.

So it is more reasonable to cache the download locally for up to 4 hours. That saves us 5–10 seconds on every run.